Table 1.

This table depicts the overall scores for the DREAM4 In Silico Size 100 Multifactorial and DREAM5 Network Inference networks using the DREAM scoring system. The iRF based algorithms outperform all three RF based algorithms. This table also shows that simply increasing the number of trees in an RF based model to match the total number of trees used in an iRF based model does not account for the overall score increase seen in iRF-LOOP. Raw AUPR and AUROC values as well as their p-values can be found in Supplementary Table A.4, Table A.5, Table A.6, Table A.7.

DREAM Challenge	Base Learner	Algorithm	Number of Trees Per Iteration	Overall Score
DREAM4	RF	GENIE3 (original)	1000	37.428
DREAM4	RF	GENIE3 (new)	1000	39.375
DREAM4	RF	GENIE3 (new)	5000	39.446
DREAM4	iRF	iRF-LOOP	1000	40.521
DREAM5	RF	GENIE3 (original)	1000	40.279
DREAM5	RF	GENIE3 (new)	1000	43.329
DREAM5	iRF	iRF-subLOOP	1000	65.466